[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAAEAJfDJp_3Z7HO41j-PhU_Cb+=gj9KQ7JarMN=dky1DaJRyWw@mail.gmail.com>
Date: Thu, 2 Dec 2021 07:38:21 -0300
From: Ezequiel Garcia <ezequiel@...guardiasur.com.ar>
To: Benjamin Gaignard <benjamin.gaignard@...labora.com>
Cc: Nicolas Dufresne <nicolas@...fresne.ca>,
Philipp Zabel <p.zabel@...gutronix.de>,
Mauro Carvalho Chehab <mchehab@...nel.org>,
Greg KH <gregkh@...uxfoundation.org>,
linux-media <linux-media@...r.kernel.org>,
"open list:ARM/Rockchip SoC..." <linux-rockchip@...ts.infradead.org>,
"open list:STAGING SUBSYSTEM" <linux-staging@...ts.linux.dev>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Collabora Kernel ML <kernel@...labora.com>
Subject: Re: [PATCH] media: hantro: Make G2/HEVC use hantro_postproc_ops
On Thu, 2 Dec 2021 at 06:26, Benjamin Gaignard
<benjamin.gaignard@...labora.com> wrote:
>
>
> Le 01/12/2021 à 21:53, Nicolas Dufresne a écrit :
> > Le mercredi 01 décembre 2021 à 15:20 -0300, Ezequiel Garcia a écrit :
> >> Hi Benjamin,
> >>
> >> On Thu, Nov 25, 2021 at 04:50:01PM +0100, Benjamin Gaignard wrote:
> >>> Use the postprocessor functions introduced by Hantro G2/VP9 codec series
> >>> and remove duplicated buffer management.
> >>> This allow Hantro G12/HEVC to produce NV12_4L4 and NV12.
> >>>
> >> Can you add the fluster score for HEVC so we confirm there are no
> >> regressions?
> >>
> >> Also, please make sure to test with the UVG set http://ultravideo.fi/,
> >> as well as testing that NV12_4L4 output converted via GStreamer's
> >> videoconvert element.
> > This is a bit unfortunate for performance, but fluster will endup testing from
> > 4L4 with how the pipeline get negotiated. Will be good to add some env at some
> > point so folks can test their CSC.
>
> fluster score is still 77/147 so no regressions.
> NV12_4L4 is the first pixel format enumerated by the driver so it is the one used
> by fluster pipeline and it is well converted into I420 by GStreamer's videoconvert.
> Do not worry NV12 is also working fine :-)
>
> I have tested Beauty, Jockey and ShakeNDry from UVG set and I see no problems.
>
Awesome!
Since we don't have any form of CI (yet?), please add the above to the
commit description
and re-submit the patch.
These drivers are a bit fragile unfortunately, and without a better
way to keep track
of testing done on each change, I feel better having this information
on the commits
themselves, specially for somewhat invasive changes like this.
Thanks and great work!
Ezequiel
Powered by blists - more mailing lists